Representing Student Argumentation as Functionally Emergent From Scientific Activity

Structure

Many studies focus on capturing the structure of arguments and supporting students in adopting components of this structure. Structural frameworks categorize students’ statements in relation to argument components and describe which components of a complete argument students use. These frameworks can be applied either to arguments written by individuals or those co-constructed in conversation. Toulmin’s (1958) field invariant form has been particularly influential in the structural descriptions adopted by scholars. According to Toulmin, a speaker who produces a claim justifies it with data, but must also show the data to be relevant to the claim by stating a warrant and backing. In addition, the speaker can specify the certainty of the claim, given the data (qualifier), and the conditions under which it would not hold (rebuttal). Many studies have used the Toulmin Argument Pattern developed by Erduran, Osborne, and their colleagues (Erduran, Simon, & Osborne, 2004; Osborne, Erduran, & Simon, 2004) to assess the quality of arguments constructed by groups of students. Levels of argument are ordered by the number of aspects of the Toulmin scheme that are present; for example, a Level 1 argument includes claims and counterclaims, while a Level 3 argument is a series of claims and counterclaims that include data, warrants, backings, and some rebuttals. Two assumptions of this approach, and of other adaptations of argument structures, are that the sophistication of an argument can be characterized by the degree to which claims are grounded and that mastery of argumentation involves using more of these argument components.

Process

Several authors have developed process-oriented frameworks that foreground the goals of argumentation, capturing the nature of discourse moves to understand how students engage in convincing each other and developing knowledge. These frameworks often use aspects of the structural frameworks reviewed above but expand them to include other kinds of discourse moves, for instance to specify whether and when students are on or off task (Clark & Sampson, 2008) are involved in management as opposed to substantive discussion (De Vries, Lund, & Baker, 2002), or are engaged in persuasive activity in contrast to explanatory or knowledge seeking activity (Berland & Reiser, 2011; Maloney & Simon, 2006). Researchers using these frameworks tend to focus on understanding which moves students make under what circumstances, as opposed to whether they construct a complete argument.

An often-cited example of a dialogic process framework is one used by De Vries et al. (2002). Students wrote explanations of a sound phenomenon, then were directed to compare explanations and resolve differences with a partner in an online environment. The authors coded students’ statements according to whether they demonstrated engagement in explanation, argumentation, problem resolution, or management. Explanatory activity involved expressing or checking on understanding, while argumentative talk included defending, attacking, or compromising. Problem resolution entailed evaluating and revising the text that partners were jointly tasked with producing. Other talk was coded as involving management of the task or social interaction. The coding scheme allowed researchers to understand students’ goals and processes as they worked on the task.

Scientific Content

In addition, many studies have tried to characterize whether students’ arguments are scientific. Structural and process-oriented frameworks are discipline-general; similar components and moves are presumed to be evident in arguments across disciplines and in everyday interaction. However, scholars have pointed out that communities develop criteria for what counts as a useful question, a possible claim, and sufficient justification (Bazerman, 1988; Blair & Johnson, 1987). There are two implications of this view: First, that the content of argument components must be evaluated, and second, that content is judged according to the norms of a discipline. For this reason, researchers have turned their attention to describing what makes an argument scientific, then supporting students to appropriate discipline-specific norms.

Many researchers identify empirical evidence as the hallmark of scientific reasoning and explore whether, and how, students use evidence in their argumentation. For example, some researchers privilege empirical data in the form of observations or measurements over appeals to other sources of justification, such as plausibility (e.g., “it seems to make sense,” “it could happen”) or authority (“That’s what the teacher told me;” D. Kuhn, 1991; McNeill, 2011; Sampson, Grooms, & Walker, 2011). However, many scholars have argued that the use of empirical data is insufficient. First, scientists actively engage theoretical commitments, including the consideration of plausibility and established facts, to construct and evaluate arguments (Koslowski, 1996). Second, data-based arguments can be flawed, if the data do not bear appropriately on the claim or are poorly interpreted. Several researchers have framed argumentation as an act of tying together theoretical claims and empirical data. For example, Kelly and Takao (2002) drew from Latour’s (1987) claim that scientific argumentation involves rhetorically stacking epistemic levels, or creating strong ties between the contingencies of experiments and theoretical claims. Similarly, Sandoval and Millwood (2005) considered how students used evidence in data charts to make claims by identifying whether students included data, pointed to data representations, or actively interpreted them.

Some schemes are designed to transcend particular contexts and domain understandings (Furtak, Hardy, Beinbrech, Shavelson, & Shemwell, 2010; McNeill, Lizotte, Krajcik, & Marx, 2006; Sampson et al., 2011), while others are closely tuned to the arguments used in a particular field. For example, McNeill characterized facility with scientific argumentation as involving the use of relevant and sufficient evidence and reasoning. Similarly, Sampson and his colleagues privileged empirical criteria, such as fit with evidence and sufficiency of evidence, and theoretical criteria that include how sufficient the explanation is and how consistent it is with other scientific ideas. They contrasted these to informal criteria, including appeals to authority, appeals to plausibility, or efforts to discredit the speaker (e.g., by arguing that another student never contributes in useful ways).

A different approach is explicitly specifying the content and inference structures characteristic of arguments within the target domain, including claims, evidence, and claim-evidence relations. Sandoval (2003) constructed a framework for assessing students’ explanations of species change in response to environmental pressure, a central form of evolutionary explanation. He privileged specific causal elements, such as environmental pressure or selective advantage, as necessary from a domain standpoint. He then described, for each causal element, what would count as a claim and how data would need to be used to support that claim. This very specific framework allowed Sandoval to understand how well students constructed an evolutionary argument, rather than a scientific argument.

While domain-free frameworks appear to possess the advantage of allowing researchers to compare students’ writing and talk across lessons or science units, on closer examination, researchers always have to deal with issues of context and domain understanding. Kelly and Takao (2002) noted that their epistemic levels representation did not assess whether the inferences the students used to move between data and theory were sensible from a disciplinary point of view; as a result, how they scored a student’s argument and how the instructor scored it often differed substantially. In addition, when authors use criteria such as relevant and sufficient evidence to assess argumentation skills, they need to translate these criteria for each question or task, for instance, by developing question-specific rubrics (McNeill, 2011; McNeill et al., 2006). As a result, comparing argumentation skill across content areas and questions is difficult. When McNeill (2011) used composite scores to represent students’ argument (combining scores for claim, evidence, and reasoning) and compared scores over the course of a school year, she found precipitous increases and drops, rather than steady development. She attributed these vacillations to the demands of the context and necessary domain knowledge, a finding supported by several other studies (Evagorou, Papanastasiou, & Osborne, 2011; von Aufschnaiter, Erduran, Osborne, & Simon, 2008).

Implications and Critiques

Structural, process-oriented, and scientific content analyses have uncovered several challenges for students. First, while students produce numerous claims over the course of argumentative activity and often justify their claims, they are less likely to provide warrants and backings (Bell & Linn, 2000; Erduran et al., 2004; Maloney & Simon, 2006; McNeill et al., 2006). Second, they usually do not spontaneously generate counterarguments and rebuttals (D. Kuhn, 1991; D. Kuhn & Udell, 2003). That is, they attend to explicating their own position, rather than identifying and addressing its weaknesses or attacking weaknesses in others’ arguments. They also do not necessarily enter instruction with an understanding of scientific norms for argumentation. For example, they have difficulty maintaining an explicit focus on the explanatory purposes of scientific argumentation (D. Kuhn & Pease, 2008), understanding what counts as a good explanation in science (Sandoval, 2003), using relevant data to support claims (D. Kuhn, Amsel, & O’Loughlin, 1988; Sandoval & Millwood, 2005), and linking claims to accepted theories (Sampson et al., 2011).

A key implication that researchers have drawn is that argument components should be taught explicitly (D. Kuhn, 1991; Osborne et al., 2004). For example, teachers can define argumentation components and provide structured opportunities for practice (McNeill & Krajcik, 2009). Students can be supported with prompts or reminders to include component parts and processes of argument (Bell & Linn, 2000; McNeill et al., 2006; Reiser et al., 2001). Desired scientific features can be supported as well. For instance, computer environments can prompt students to apply causal reasoning by repeatedly asking them to identify whether a factor “makes a difference” or “makes no difference” (Kuhn & Pease, 2008). In some cases, students are prompted to use relevant evidence, for example by receiving reminders to avoid including measurements that are not properties (i.e., mass or volume) in their written explanations (McNeill & Krajcik, 2009). These forms of support are often conceptualized as scaffolds that allow students to engage in argumentation before they are able to do so independently, and then are faded out as students develop proficiency (McNeill et al., 2006; Quintana et al., 2004).

These approaches constitute a bet that supporting students to use scientific tools in the ways that scientists use them will allow students access to scientific activity systems. That is, using scientific argumentation forms will orient students to the knowledge-building and persuasive goals of science and support them in making progress toward those goals. The introduction of tasks and structures that promote argumentation has helped us make important progress toward these goals. Organizing classrooms around argumentation encourages students to interact directly with each other, as opposed to directing conversation exclusively to the teacher, who, in turn, evaluates comments (Berland, 2011; Duschl & Osborne, 2002; Martin & Hand, 2009). Making argumentation structures visible to students encourages them to make their ideas explicit, promoting the elaboration, connection, and consolidation of scientific understandings (Bell & Linn, 2000; Chin & Osborne, 2010; de Lima Tavares, Jiménez-Aleixandre, & Mortimer, 2010; Keys, Hand, Prain, & Collins, 1999; von Aufschnaiter et al., 2008). Some studies have shown that, over time, students appear to use more of the desired components and scientific norms in their argumentation (McNeill et al., 2006; Osborne et al., 2004; Sampson, Enderle, Grooms, & Witte, 2013).

In spite of the way that these studies have illuminated how students tend to think about argumentation and shown how rare it is in classrooms without explicit support, there have also been critiques of these approaches, on the grounds that they are not sufficient for allowing students adequate access to the activity systems within which scientific argumentation operates. First, several authors have argued that argument structures may fail to adequately capture the varied forms of activity that support the development and defense of scientific knowledge claims. Researchers have yet to develop agreement on exactly what constitutes a scientific argument and whether it can be described outside of a particular domain. Also, structural- and content-based representations typically leave out activity that is conjectural, nonverbal, or focused on material outcomes, such as the development of an instrument (Bricker & Bell, 2008; McDonald & Kelly, 2012; Shemwell & Furtak, 2010). In addition, it appears that students can take up the forms of scientific argument without understanding their purpose, using taught structures as tools to complete lessons, rather than to engage in scientific activity (Berland & Reiser, 2011; De Vries et al., 2002; D. Kuhn & Pease, 2008). Finally, specifying the desired structure and content of scientific arguments can cause insensitivity to the context within which argumentation occurs. For example, it is plausible that students fail to warrant claims not because they do not know how to, but because they feel they do not need to (Kelly, Druker, & Chen, 1998; Sandoval & Millwood, 2005). Both domain-general and disciplinary content-based structures are distillations of one aspect of scientists’ work—final-form arguments of the type published in scientific journals—but appear not to constitute a sufficient support for students to participate meaningfully in working together to build and revise scientific knowledge.

These problems gain additional focus when they are considered from a perspective on the nature of social practice. Although practices manifest as routine structures, their forms are interactionally accomplished in communities to facilitate a shared enterprise (Wenger, 1998). Wenger described routinized, describable practices as “reifications” that support community activity, but cautioned that these routines are “double-edged” in the sense that they erase the negotiated activity that originally gave them meaning (p. 61). Engaging students in doing what scientists do when they argue—for example, using evidence or making coherent links between disciplinary ideas—constitutes an initiation into reifications of scientific argumentation. However, students might not understand the purpose of these forms of talk in the larger scientific enterprise. Therefore, they may be able to engage in argumentation without experiencing it as a knowledge-building practice for themselves in their scientific work. By developing stable descriptions of argumentation and using them to inform assessments and supports, we risk lifting the practice from the contexts that lend it meaning.

For this reason, researchers have advocated for shifting attention to the contexts within which argumentation takes place and establishing classroom environments in which students experience a need for the practice (Berland & Reiser, 2011; Sampson et al., 2011). In the next section, I consider the implications of this goal. I describe important differences in the activity systems within which scientists and students operate and explore ways that researchers have attempted to bring these aspects of activity systems into alignment with each other.

Scientific Enterprise

Philosophers and sociologists of science conceptualize science as a modeling enterprise characterized by significant uncertainty (Giere, 1990; Longino, 2002; Pickering, 1995). The view of science as modeling is one that is increasingly advanced by the science education and learning sciences communities (Lehrer & Schauble, 2006; Schwarz et al., 2009; Stewart, Hafner, Johnson, & Finkel, 1992; Windschitl, Thompson, & Braaten, 2008). The modeling perspective highlights the representational and material objects of scientific activity. Scientists produce abstract, predictive representations of the world (models) in the form of explanations, equations, and physical instantiations such as the Bohr model of the atom (Giere, 1990). Models are judged by their perceived fit with the world, the actions that they allow scientists to take, and the work that they make possible; much of scientific activity consists of deepening and expanding models or constellations of models (Giere, 1990; T. Kuhn, 1996; Longino, 2002). Modeling also has a highly material sense, in that scientists have to harness and interpret what Pickering (1995) referred to as nature’s agency. Because nature does not always talk in ways we can easily understand, scientists attempt to “capture, seduce, download, recruit, enroll, or materialize” its agency (Pickering, 1995, p. 7). To do so, they make models of phenomena in the forms of experiments, machines, stem cell cultures, or animal models, and manipulate them to construct and test arguments.

Modeling is permeated with uncertainty; much of science is concerned with managing this uncertainty. Theories and explanations involve hypothesized events and unseen processes; they are by nature conjectural and revisable (Windschitl et al., 2008). But even the empirical observations that support these accounts are fundamentally uncertain; rather than merely using evidence, scientists make it. Phenomena must undergo processes of construal to become objects of observation, reflection, and communication (Gooding, 1990). Instruments and experimental protocols are developed over time and are subject to significant scientific energy (Knorr Cetina, 1999; Nersessian & Patton, 2009; Shapin & Schaffer, 1985). Therefore, concepts and the practices that produce them are intricately bound together. Pickering (1995) argued that facts are produced through stabilizations of material work consisting of the design and revision of machines and measures, interpretive accounts of material results, and phenomenal accounts that assign meaning or status to the natural objects that are the subject of their work. Each of these three aspects of practice is fundamentally uncertain; their alignment creates stable objects that can be taken as knowledge.

Scientific Communities

Recent accounts frame the scientific community as the locus for the stabilization of model, practice, and machine (Latour, 1987; Longino, 2002; Pera, 1994). Pera (1994) conceptualized the typical, science as method, view as a play with two actors: the scientist, who proposes; and nature, who answers with a ringing “yes” or “no.” In contrast, he argued that a more appropriate portrayal involves three actors: the proposer; nature, who speaks; and the community, who determines what is to be taken as the official interpretation of nature’s voice. From this point of view, participants in a scientific community alternate between two roles; they construct and critique the links between conceptual accounts and the material activity that supports them (see Michael Ford’s work, e.g., 2006, 2008, for a detailed characterization of construction and critique). Latour and Woolgar (1986) illustrated how these roles stabilize what scientists take as known. They argued that facts emerge as contingent results in laboratories; are inscribed and made visible to others, become subject to debate over methods, theory, and even the qualities of the researchers who created them, and are gradually recruited into larger networks as they become useful to others. Statements that are taken as fact are stripped of the modal qualifiers that were integral in their creation; that is, the way in which they became known is made invisible. Over time, these facts become integrated into the discipline, reified in machines and instruments, and, in turn, support the process of developing new facts.

Longino (1994) argued that by treating theories as models and studying the communities in which models are developed and contested, we can develop an account of why science has “advanced.” She wrote,

If S₁ . . . S_n are members of an epistemic community [C], W is some real-world system or portion of a real-world system, and M is a model (of that system) then S1 . . . S_n know W as M if and only if

This description, which has been analyzed elsewhere (e.g., Kelly, 2008), embeds scientific knowing and argumentation within community activity. The first two criteria frame knowledge as a model that allows members of the community “to act” and “satisfy their goals.” This is a representational view that also privileges the performative acts of modeling and the tangible results of knowing. The third criterion describes characteristics of the community that allow it to develop knowledge; it provides opportunities for interaction (iiia), involves the uptake of critique (for instance, in that scientists engage in further construction of ideas and procedures in response to critique), and includes different perspectives that are accorded enough authority to be considered (iiid). The final criterion, public standards (iiic), indicates that shared norms and values (e.g., empirical adequacy, breadth, fulfilling technical needs, and coherence) facilitate the development of shared knowledge.

It is beyond the scope of this article to fully characterize the complex divisions of labor and power or the values that characterize scientific communities. Professional activity is organized in labs in which participants carry out different projects, develop different expertise, and sometimes work across disciplinary boundaries (Gallison, 1999; Nersessian & Patton, 2009). Labs include oldtimers, with established access to machines, ideas, and other resources and newcomers, whose access is more circumscribed. Some labs are able to amass physical and political resources that effectively silence others’ ideas (Latour, 1987). In addition, scholars increasingly argue that it is impossible to characterize a single set of norms that guide scientific activity or scientific argumentation. Instead, scientific communities are guided by distinctly different norms for what counts as knowledge and for acceptable warranting practices (Gallison, 1999; Knorr Cetina, 1999; Mayr, 2004).

It is important to recognize, however, that roles and norms are not static or pre-determined by the greater scientific enterprise; instead, they are developed in scientific communities working around shared questions, models, and instruments. For example, the notion of empirical observation, which has become central to physical science, was promoted by Robert Boyle, and contested by other scientists, as Boyle sought to use a mechanical device (the air pump) to develop pneumatic theories. Shapin and Schaffer (1985) showed that Boyle’s attempts to convince the community to accept his ideas involved negotiating standards for replication and witnessing as well as what counted as public activity and what roles were available for those who wanted to participate in making scientific knowledge. Developments like these give rise to different epistemic cultures (Knorr Cetina, 1999) characterized by particular roles, problems, machines, and representations.

Reconsidering Argumentation as a Tool Within a Scientific Activity System

From this point of view, the need to argue emerges from the inherent uncertainty of the modeling endeavor. Because science involves stabilizing material, interpretive, and phenomenal accounts by bringing them in line with each other, a theory or model can be evaluated, and attacked, from any of these points of view. Latour and Woolgar (1986) as well as Latour (1987), showed how disputes involve reintroducing modalities and tying supposed “knowledge” back to the conditions in which it was created, framing it as an artifact of activity. Argumentation focuses on providing, teasing apart, and refining the complex web connecting the material, particular world with abstract statements. It works not only on claims, but the entire machinery that supports claim-making, including measures, instruments, experiments, inscriptions, and concepts.

If knowledge and the machinery that supports knowledge construction emerge from argument, how do forms and modes of argument come to be accepted and used in particular communities? They themselves emerge from the scientific activity of the community and shift with new machines, representations, and shared knowledge. Consider, for example, the experimental article, a central persuasive tool of scientific communities. Over time, what constitutes an experiment and the appropriate way to use an experiment to make an argument have undergone immense change (Bazerman, 1988). Originally, experimental articles presented experiments as any manipulation of nature (e.g., methods for coloring marble). As articles became subject to debate, it became increasingly apparent that understanding what was happening during the manipulations, and why, was often problematic. The experiment became a way of solving a puzzle or supporting a claim. Methods of experiment were then scrutinized, leading to development in the careful reporting of procedure, the use of multiple trials, and deliberate attempts to eliminate alternative explanations. In the 19th and 20th centuries, experiments were positioned as ways to explore and further theories, and new criteria for arguments, for instance, of demonstrating significance for theoretical advancement, became codified in the structure of the papers. Written arguments, then, have evolved from descriptions of action to intricately constructed explications of theory, precise and repeatable procedures, and the justified interpretation of results. This trajectory is intimately related to trajectories in other practices, for instance the experiment (Bazerman, 1988).

The final-form argument as presented in conferences or articles can be conceived as a codified set of strategies for cleaning up the stabilization process so that others accept the proffered account. Gooding (1990) traced the many forms of reconstruction that support the arguments reported in scientific journals, including cognitive reconstructions that allow scientists to justify actions in real time (e.g., in-the-moment experimental decisions), demonstrative reconstructions that allow them to describe what they did to others, and methodological reconstructions that relate arguments to accepted methods. Reconstructions rest on internalizations of critical voices that seek to undermine links between concepts, models, and material work. These internalizations can be both prospective, in that the writer anticipates possible objections (Billig, 1996), and retrospective, in that they are responses to criticisms already made (Bazerman, 1988). It is at this stage that the argument can be neatly parsed into constructs such as theory, causal account, and empirical evidence. However, the form in which it is presented purposefully black boxes much of the work needed to develop the argument.

Conceptualizing science as a modeling activity positions argumentation as the practice that problematizes and stabilizes both what communities know, or the models they use, and their means of knowing. Participation in argumentation can be framed and assessed as participation in problematizing and stabilizing a community’s material and representational activity. The final-form argument presented in journal articles is one aspect of this process, but it erases the historical and personal activity that supports it. The framing of argumentation presented here places our focus not on argumentation as an isolated, static practice but on scientific activity more generally. I now turn to understanding how this framing might inform our work with students.

Classroom Enterprise

To embed argumentation in classroom communities, it is important to consider students’ enterprise and the problem-space within which they are asked to engage in argumentation. To this end, researchers have noted that students’ and teachers’ enterprise differs significantly from scientists’ and have developed supports for students to adopt knowledge-building and persuasive goals. However, to date, here has been less emphasis on engaging students with uncertain objects and theorizing the role of uncertainty in situating the development of a modeling enterprise.

The activity of schooling privileges the performance of knowledge or skills that are taken as certain, in that they are specified in curriculum documents and assessments and are known by the teacher. As a result, student motives (and thus goals that are realized in students’ actions) typically involve performing competence; for example, raising one’s hand to answer a question correctly or getting a good grade on a test. Therefore, argumentation structures that are introduced to students are likely taken up for the goals prominent in this activity system, those of performing knowledge and skill (e.g., doing what the teacher tells you). As this issue has become evident in the research on student argumentation, researchers have focused on understanding how students adopt persuasive and knowledge-building goals, as opposed to school-like goals such as completing assignments and pleasing the teacher (Berland & Reiser, 2011; Jimenez-Aleixandre, Rodriguez, & Duschl, 2000).

This research has shown that many students do not adopt these scientific goals, even when they use argumentation processes or structures. For example, Kuhn and Pease (2008) asked students to include a statement of the purpose of inquiry in a written research report. Although 17 of the 18 students supported their claims with evidence and a majority correctly described aspects of the procedure that allowed them to collect conclusive evidence, only half identified that the purpose of their investigation was to find out which of several factors mattered. Berland and Reiser (2011) used the coding of discourse moves in a comparative case study to demonstrate that one classroom appropriated a goal of individual sense-making while another classroom was characterized by a stronger goal of persuasion. They claimed that the two goals may be in tension in classrooms and that both teachers and students can experience difficulty in reconciling them. From this point of view, engaging students in collective sense-making, which necessarily entails both persuading others and seeking to make sense of ideas, is challenging. However, other studies have shown students taking up these goals as they argue (Keys et al., 1999; Naylor, Keogh, & Downing, 2007). These findings have been used to suggest that students’ goals are deeply tied to classroom tasks and participation structures (Berland & Hammer, 2012).

Researchers have used several different approaches to design learning environments that support students to adopt knowledge-building and persuasive goals. Some studies have experimented with how goals might be explicitly built into the directions for argumentation tasks. For example, Garcia-Mila, Gilabert, Erduran, and Felton (2013) compared how task directions to persuade a partner as opposed to reach a reasoned consensus affected students’ argumentation. They found that students’ argumentation activity was richer when they were provided, and presumably adopted, knowledge-building goals. Another common strategy is to implement activity structures that bring students who disagree into contact with each other, so that they need to share ideas and try to reach a new understanding together (Berland & Reiser, 2011; Clark & Sampson, 2008; De Vries et al., 2002).

Other researchers have shifted their attention from students’ in-the-moment goals for practice to the enterprise within which argumentation is situated. For example, Berland and Hammer (2012) stressed the importance of teachers framing the work within which argumentation is embedded as concerned with finding out something. From this point of view, for students to engage in goals such as convincing others and building knowledge, they must conceptualize their wider activity as concerned with finding something out, rather than demonstrating a particular kind of discourse skill or structure. Another approach is embed argumentation, and sometimes scientific activity more generally, in enterprises that students already recognize as consequential, for example, discussing social issues (D. Kuhn, 2010) or testing the quality of drinking water in their school (Rosebery, Warren, & Conant, 1992). The assumption here is that students will more easily take up knowledge-building and persuasive goals and apply resources to meet these goals in these contexts, and that they can then be supported to see these goals and attendant strategies as useful for scientific argumentation.

Another aspect of the scientific enterprise that warrants further consideration is the fundamental uncertainty of the material and representational objects that scientists work with. If argumentation is functionally emergent from the scientific problem-space, it should matter for students’ argumentation both whether they are engaged with uncertain material and representational objects and how their activity is framed. In his review of 54 argumentation interventions, Cavagnetto (2010) considered whether students participated in representational and material activity, using Ford’s (2008a) conceptualization of getting nature to speak (here, material activity) and representing nature’s voice (representational activity). Cavagnetto found that 32% of argumentation studies engaged students in neither material nor representational work, 50% included representational but not material activity and that only 18% (10 articles) engaged students in specifying problems, posing questions, designing and conducting investigations, or explaining investigation choices to others. Several recent papers have argued for the importance of engaging students in the material work of science as a support for argumentation (Katchevich, Hofstein, & Mamlok-Naaman, 2013; Sampson et al., 2013).

In addition to engaging students with these aspects of science, it is also important to consider how material and representational work are presented. If students are to participate in stabilizing their own scientific work, they must experience enough uncertainty that both knowledge and the means of knowing are legitimately problematic. Studies that have examined the role of uncertainty suggest that it is productive for supporting argumentation practice. For example, Watson, Swain, and McRobbie (2004) examined students’ argumentation as they made paper chains, tested to see if a particular factor mattered for the strength of the chain, and represented their results. However, all methodological choices were specified by their previous work or intervention by the teacher. Students produced little argumentation; they generally sought neither to justify nor to warrant claims. The authors attributed the lack of argumentation to the routinized nature of inquiry activity; in other words, directions and expectations constrained opportunities and there was little uncertainty for students to resolve. Palincsar, Anderson, and David (1993) compared two iterations of one scientific task, demonstrating that the choices the designers made, for instance specifying (or not) for students what solvent to use and what constituted a fair test, were instrumental in determining the opportunities for students to engage in debate. When they were encouraged to grapple with uncertainty, students brought significant conceptual resources to bear on designing their investigation and engaged in prolonged discussion before developing a sophisticated measure of the investigation’s target outcome.

Analyses like those conducted by Palinscar et al. (1993)—that is, those that focus on the aspects of argumentation available to students based on the constraints and affordances of instructional tasks—are rare. However, these forms of analysis might be essential to support student argumentation. Educators often fear that uncertainty makes argumentation practice more difficult for students, and therefore tend to present evidence in the form of statements or limited data sets, particularly in their work with younger students (Berland & McNeill, 2010; Chin & Osborne, 2010). They are more likely to build uncertainty into students’ representational activity, for example by using open-ended tasks with at least two possible plausible solutions (Berland & Reiser, 2011; Jimenez-Aleixandre, 2008) or providing data that are both relevant and irrelevant (Reiser et al., 2001). Fewer studies build uncertainty into students’ material activity. In addition, studies do not generally specify how uncertainty is built in to the design, other than noting that multiple plausible explanations are possible. However, if we expect argumentation to emerge from uncertain activity, it is likely to matter what about students’ work is uncertain, for example, which claim to select, which evidence to use, how to interpret evidence, or how to make evidence. It is worthwhile to analyze, in any instructional sequence, what decisions are available to students, where productive uncertainty might be conjectured to appear, and what means of resolving uncertainty students can take up. These are the most likely sites for argumentation and for the development of an understanding of the scientific enterprise.

It is important to note here that the focus on performance and the typical absence of uncertainty in classrooms exist for reasons that extend beyond the classroom and the curriculum designs that guide it. Therefore, there is likely to be significant resistance to the some of the shifts involved in more adequately representing the scientific enterprise in classrooms. I discuss some of these challenges in the next section and then return to them in the article’s final sections.

Classroom Communities

In this section, I examine important differences between scientific and classroom communities that are consequential for the uptake and use of argumentation, then explore how researchers have addressed these differences. As described above, the scientific community comprises the individuals who construct and critique shared objects and attempt to establish knowledge together. For students to participate in meaningful uses of argumentation, the practice, therefore, must be situated in their classroom community and will be influenced by divisions of labor, roles, and norms available in this setting.

First, the roles and power relations of a classroom differ in consequential ways from those available in scientific activity. These differences have been the focus of substantial work within the research on scientific argumentation. Researchers have emphasized that argumentation demands a shift in the role of the teacher, who must move from traditional recall or yes/no forms of questioning to seeking student opinions, eliciting or elaborating ideas, and framing students’ activity as about building knowledge and consensus (Berland & Hammer, 2012; Martin & Hand, 2009). In addition, there has been an emphasis on providing students with the authority to find something out, as opposed to providing the correct answer (Chin & Teou, 2009; Duschl & Osborne, 2002; Jimenez-Aleixandre, 2008). Third, researchers have developed activity structures that position students as an audience for each other’s work and encourage revision of ideas in response to critique. For example, students might engage in peer review of each other’s written work (Sampson et al., 2013; Sampson et al., 2011) or participate in scientific research meetings in which they share their work and are questioned or critiqued by peers (Ford, 2008b; Herrenkohl, Palincsar, DeWater, & Kawasaki, 1999). These recommendations support a shift in the typical roles available in the classroom, from positioning the teacher as knower and judger and student as producer of the right answers, toward supporting a knowledge-building community in which students both construct and critique ideas.

However, when we shift roles so that students are accountable to each other rather than directing comments to the teacher, another aspect of communities, the norms that guide practices, become consequential. Typical representations of argument are usually described from a normative view of the practice or discipline as enacted in expert contexts; that is, they represent a description of what the practice should look like or sound like. However, these features might not represent the aspects of argumentation needed in the classroom community. For example, one common assumption is that facility with argumentation involves explicitly using all components of the argumentation structure. In contrast, Kelly et al. (1998) found that students working together on open-ended electricity problems often failed to warrant evidence because they shared a common understanding of problems, materials, and potential data sources that rendered a warrant unnecessary. The authors were able to trace categories of events in which a need for warrants arose, including questions from other students and the presence of anomalous data. Similarly, Clark and Sampson (2008) found that students were more likely to warrant their claims when they were responding to opposition from other students. These findings raise serious questions about whether it is fair to assume that arguments lacking grounding are less sophisticated; instead, they might reflect an arguer’s awareness of the common ground shared by his or her audience.

In addition, descriptions of scientific argumentation typically characterize the arguments that scientists make, for example, their causal nature, the role of empirical data, or the need to interpret figures. But it is not clear that these features would be necessary to convince students’ audience—their classmates, who likely do not initially hold these disciplinary values, or their teachers, who already “know the answer.” As Sandoval and Millwood (2005) pointed out, it is difficult to interpret an aspect of scientific argumentation that students fail to incorporate in their argument, for example, interpreting a graph, as indicating something that they do not understand or as an assessment of a shared common ground, in the sense that they believe their teacher knows how to interpret the inscription. Conflicts exist between normative models of argumentation, which privilege particular structures or content used in expert practice, and dialogic models, which focus on students constructing and critiquing knowledge together.

Scientific activity systems do involve unequal power distributions that are likely productive in initiating newcomers into scientific norms. Science is carried out by newcomers and oldtimers working together, where newcomers take on roles as apprentices and build knowledge by taking up normative practices. In addition, labs are characterized by distributed expertise, in that scientists work on different methods and ideas. How to develop these hybrid structures in classrooms, where students are both empowered to participate in scientific activity and supported to use powerful disciplinary practices and ideas, is an enduring question (J. S. Brown, Collins, & Duguid, 1989; Engle & Conant, 2002; Ford & Wargo, 2012). In work specific to argumentation, researchers have examined teacher moves that hold students accountable to each other, engage them with disciplinary ideas and evidence, and highlight relations between their work and what scientists do (Forman & Ford, 2013; Herrenkohl & Cornelius, 2013). In each of these examples, the teacher directs his or her leadership role to developing a community and helping students see the power of disciplinary norms for supporting productive construction and critique. In other work, researchers have invited practicing scientists into the classroom, so that students can be questioned and critiqued by experts modeling disciplinary norms (Jurow, Hall, & Ma, 2008; Meyer, 2014). However, there is still much to learn about how scientific norms and values are best introduced to students in contexts where they are empowered to use them as well as how they begin to take on these values and hold each other accountable to them.

Argumentation as a Tool for Students

Given the differences in the activity systems within which scientists and students typically enact argumentation, as described above, it is sensible that students adopt new structures for the purpose of pseudoargumentation, that is, to please the teacher or do well on a test (Berland & Hammer, 2012). They are effectively using these structures as a tool to accomplish their goals in the schooling enterprise as they experience it. The question arises, then, how we might better help students recruit argumentation as a tool that has similar meanings and uses to those it has in scientific activity systems. Here, I explore three suggestions for supporting a more productive alignment of argumentation in scientific and classroom activity systems: positioning argumentation as integral to students’ scientific activity, understanding how it helps them develop both what they know and how they know it, and considering how students’ uses of argumentation in enterprises outside of school can serve as a resource for tool use in this different setting.

First, researchers need to develop classroom activity systems in which students’ argumentation can serve an integral role. To do so, we will need to refine our understanding of what makes argumentation integral to scientific activity in classrooms (Cavagnetto & Hand, 2012). Argumentation activities that span a single lesson, are unrelated to the subject matter knowledge that students are in the process of developing, or serve simply as culminating activities are unlikely to allow students access to the purposes that argumentation serves in scientific activity. Cavagnetto (2010) concluded that 63% of argumentation studies fell into one of these categories; only 37% involved students in argumentation activity that was integral to their inquiry. However, in Cavagnetto’s review, even when argumentation was embedded in activity, it is not always clear what it meant for it to be integral. In some studies (e.g., Naylor et al., 2007) students participated in argumentation to promote the need to investigate. In others, argumentation was conceptualized as an opportunity to revise explanations based on data, but no additional empirical work was pursued (e.g., Berland & Reiser, 2011; Clark & Sampson, 2008). These uses of argumentation might engage students in problematizing each other’s ideas and possibly methods, but are less likely to support the stabilization of shared warrants and investigative procedures.

I conceptualize argumentation as integral to students’ activity if it allows them to contest and agree on how they might know something, in turn, stabilizing particular models or inspiring new questions and models. For example, Herrenkohl and her colleagues (Herrenkohl & Mertl, 2010; Herrenkohl et al., 1999) involved students in developing theories that explained the data they were collecting. The class kept a public chart of theories, which they revised as they collected new information. In order to agree on what to include in the chart, they participated in protracted argumentation to define what would count as a theory, explored the mechanistic qualities of theories, and decided that their theories had to specify not only what happened, but also why it happened. In turn, this definition was integral to their further work because it allowed them to distinguish between theories that were worth exploring further and those that were underspecified and less useful for future work. Several studies have shown that students, from a young age, can recognize uncertainty, engage in productive discourse about it, and develop ways to deal with it in their ongoing scientific activity (e.g., Ford, 2005; Lehrer, Schauble, & Lucas, 2008; Passmore & Svoboda, 2012).

To shift argumentation from a procedure conducted after scientific work has been done, or conducted in the absence of scientific work, and instead embed it in the scientific enterprise, we will also need to develop tools that allow teachers and researchers to understand how students participate in this work. One issue is that typical structures generally treat the explanatory claim as the focus of all argumentation, and as a result, flatten the work that goes into coping with uncertainty and making evidence into a warrants or reasoning component that is explicitly related to that claim. These exchanges are usually represented as a few lines of dialogue, in contrast to the extended conversations that take place in scientific activity.

Some lines of research are beginning to focus on the relationship of the structures that support final-form arguments and the forms of talk that students engage in throughout their scientific activity. For example, the Science Writing Heuristic (SWH; Cavagnetto, Hand, & Norton-Meier, 2010; Keys et al., 1999) is a paired template for teacher and student activity that extends over the course of students’ initial thinking about a phenomenon, engagement in laboratory activities, analysis of data, and development of explanations grounded in evidence. Researchers promoting the SWH have stressed the importance of attending to how data are made into evidence and treating this as an ongoing process, rather than a component of a final-form argument (Cavagnetto & Hand, 2012; Keys et al., 1999). Recently, they have turned their attention to understanding how the heuristic supports classroom interaction (Cavagnetto et al., 2010) and what students attend to as they construct and critique knowledge claims, including the quality of the evidence provided and its relationship to the claim made (Chen, 2011).

Studies of another model, Argument Driven Inquiry (Sampson et al., 2013; Walker & Sampson, 2013) have taken a similar approach to the SWH work, embedding argumentation in inquiry and representing both students’ talk and their writing. To date, most published research has focused on students’ written work (e.g., Sampson et al., 2013). Sampson, Enderle, and Walker (2012) have also developed a protocol for examining argumentation talk, Assessment of Scientific Argumentation in the Classroom, that examines talk along three dimensions: conceptual, epistemic, and social. The social and cognitive dimensions indicate how students interact and whether they view ideas as something to be considered and revised. The indicators for engagement in epistemic work approximate the evidence-making aspects of scientific activity, in that they focus on representing whether participants “examined the relevance, coherence, and sufficiency of the evidence” and “evaluated how the available data was interpreted or the method used to gather the data” (Sampson et al., 2012, p. 245). The one published study using this rubric (Walker & Sampson, 2013) showed some increase in students’ use of these features over time, but also suggested that these features were deeply tied to aspects of the task, such as what was available to argue about.

These approaches embed argumentation within a scientific enterprise, attend to whether students engage with dialogic processes (e.g., putting forward ideas, questioning others, agreeing or disagreeing) and describe the scientific targets of these dialogic moves, including whether students begin to see questions, concepts, or experimental procedures as something to agree and disagree about. They focus less on the structure of talk and more on what is considered as a target for agreement, disagreement, and progress. Therefore, they are more closely aligned with the framing of argumentation presented in this article, which emphasizes participation in contesting and stabilizing both what is known and how it is known.

Further research might include explicitly designing instruction to support students in iteratively developing and contesting the grounds for knowing, for instance, what would count as an acceptable form of data, whether a particular instrument can be used and how, which interpretation of a graph is best, or whether a student-constructed data table adequately represents findings. To understand argumentation, one could then specify first what parts of activity were subject to variability and open to contest, and then describe how students participate in contesting and stabilizing them. For example, Radinsky (2008) explored how groups of students negotiated and co-constructed the meanings and purposes of GIS database inscriptions. He examined how references to data, to real-world objects such as oceans and particular continents, and to domain concepts (e.g., buckling or plate) were used in sense-making and related to each other in contradictions, questions, and claims. He identified instances of students constructing and challenging data images that supported their claims, heuristics for identifying instances of disciplinary concepts, and shared understandings of concepts. In this study, students debated and developed both geological concepts and shared ways of constructing and interpreting data.

In addition, engaging students in social processes with the purpose of developing scientific ideas and evidence might allow them to leverage tools from other activity systems to participate in the scientific enterprise. This perspective supports inquiry into how children’s practices can be framed as a resource for engaging in scientific argumentation (Bricker & Bell, 2008). Research has shown that argumentation is central to children’s out-of-school concerns, including negotiating social order, establishing identity, and getting what they want (Goodwin & Goodwin, 1987; Kyratzis, 2004). In these activity systems, students competently argue using strategies that include threats, intonation, format-tying, attacking character, telling stories, questioning justifications, and probing cause-effect logic and alternative explanations (Bricker & Bell, 2012; Goodwin & Goodwin, 1987; Ochs & Taylor, 1992). While these arguments do not sound scientific, they reveal important continuities with the tools needed to participate in scientific argumentation: children adjust their argumentation strategies to particular situations and co-participants, draw on a wide range of resources to convince others, and construct social organization through agonistic exchange.

Some educators have successfully capitalized on these practices to support disciplinary activity. For example, Hudicourt-Barnes (2003) invited students to participate in a Haitian–Creole diskisyon practice around their scientific activity. In these discussions, participants take up the role of theoretician, challenger, and audience. In one example presented in the article, students challenged the basis of a claim about snail reproduction, arguing that not enough time had passed for the snails to produce multiple generations and identifying aspects of snail life cycles (e.g., time to maturation and reproduction) that were in question. In another, they questioned the basis of the comparison a student was making and proposed an experiment to verify or disprove the claim. In these examples, students leveraged social practices that they already found sensible to engage in providing, questioning, and refining evidence. Warren, Ballenger, Ogonowski, Rosebery, and Hudicourt-Barnes (2001) found that another group of students supported to use diskisyon applied informal practices of humor, teasing, and acting to debate the differences between growth and change, building a nuanced understanding of metamorphosis. Varelas et al. (2007) traced how urban second and third graders used informal and scientific language to debate whether objects were solids and liquids, contesting definitions and even units of analysis (e.g., whether a grain of sand or a vial of sand was the correct unit for examining whether the object held its shape). Each of these studies showed how students’ practices of agreeing, disagreeing, and questioning were leveraged to support the discussion of ambiguous ideas and evidence.

To date, most accounts of argumentation emphasize how students support knowledge claims rather than how they contest and develop their means for knowing. Therefore, most studies focus on how students use evidence, rather than how they participate in making it. As a result, a key function of argumentation is underrepresented in the literature. This focus has translated into representations of scientific argumentation, which tend to collapse conversation about variables, measures, experiments, and data representations and examine argumentation after empirical work has been completed. This approach presents a simplification of argumentation that approximates reconstructions of activity that occur after investigations have been concluded and are made by experts who have internalized the voices of other community members. It represents a developmental bet that this is a useful introduction into argumentation for students, who might over time complicate the argument structures that they are taught, apply them to increasingly complex questions, and use argumentation to participate in a greater proportion of the scientific enterprise. However, there is growing evidence that another approach might have significant affordances: positioning argumentation as a social process for which students bring in significant resources, then helping them see the scientific enterprise as worth applying those resources to. Both of these perspectives position argumentation as a knowledge-building tool, but they represent different approaches to aligning students’ and scientists’ activity systems.

Entries Into Practice

The first question to address is what constitutes a useful introduction for students into argumentation practice; that is, what aspects of a scientific activity system do we introduce to support entry into scientific argumentation? For each aspect of activity systems, researchers might choose to begin with students’ resources, to introduce normative aspects of the scientific activity system to shift practice, or to invoke classroom structures that support teaching and learning. Below, I outline some potential choices that will require justification and further research.

Development

From the perspective presented here, it is important to attend to the development and adaptation, rather than the adoption, of practice. If practices are considered to be interactively constituted in activity (Rouse, 1996; Wenger, 1998), it is impossible to expect that the practice as introduced to students is identical to the practice as taken up and used by them over time, especially if it is contributing to the development of a classroom epistemic culture. A first criterion for attending to the development of practice is lengthening the time-span of argumentation studies, because shifts in community do not occur within a day or week. While many studies focus on interaction during single episodes, several recent studies have described argumentation over longer time periods (e.g., Chen, 2011; McNeill, 2011; Ryu & Sandoval, 2012; Sampson et al., 2013; Tang & Coffey, 2010). From the perspective adopted here, three related forms of development are important to understand: norms for the practice, shifts in the scientific activity system, and changes in an individual’s facility with argumentation.

Supports for the Development of Practice

Finally, we will need to understand how to support students and teachers to participate in the development of scientific activity systems. That is, we must understand how to set up learning environments where teachers and students use argumentation as a productive tool to develop shared knowledge, shared knowledge-building practices, sophisticated norms for argumentation practice, and even new forms of social relations. While science might be characterized by open-endedness and uncertainty, schools are not. They privilege a telos, which is codified in lesson plans, textbooks, and assessments. Therefore, if we want to support teachers to engage students in argumentation, we must find ways to help them know where they are going and provide them the tools to get there.

Argument structures are a reification that provides an instructional goal. As such, they are often highlighted in work with teachers (McNeill & Krajcik, 2009; Simon, Erduran, & Osborne, 2006). However, this emphasis can easily be translated into framing teaching as introducing these structures to students, defining their component parts, and providing feedback on students’ use of those parts and conceptualizing assessment as testing which components students use. Numerous studies have shown that argumentation structures are insufficient for supporting teachers to make the shifts that this article has suggested are important (Berland & Reiser, 2011; Osborne, Simon, Christodoulou, Howell-Richardson, & Richardson, 2013; Simon, Richardson, & Amos, 2012).

If we want, instead, to support teachers to focus on developing epistemic cultures in classrooms, we will need to develop new knowledge, curriculum structures, and forms of support. We will need to specify the ideas and material practices that are important for a topic or investigation and understand how to make these ideas and practices uncertain for students. We will need to support teachers to position students as a community of knowledge and evidence makers, to recognize productive moments of uncertainty, and to use teaching moves to help students to discuss and develop target ideas and practices. It is likely that this work will require content-specific descriptions of target ideas, practices, ways for students to engage in productive discourse around these ideas and practices, and the shifts that constitute progress. Domain-specific learning trajectories and supports of this kind have been successful in supporting the development of arithmetic strategies (Carpenter, Fennema, Peterson, Chiang, & Loef, 1989; Fennema et al., 1996) and understanding of statistics (Cobb, McClain, & Gravemeijer, 2003; Lehrer, Kim, & Schauble, 2007; Lehrer & Schauble, 2004) in learning environments where students share and compare strategies.